Hydraulic engineering safety monitoring method and device and storage medium
By using drones to carry target water conservancy project safety monitoring models, combined with deep learning and feature extraction technologies, the problem of low efficiency in manual supervision during water conservancy project construction has been solved, achieving efficient and accurate safety monitoring, reducing human resource costs and regulatory blind spots.
Patent Information
- Application Number
- CN202511663967.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-13
- Publication Date
- 2026-02-17
AI Technical Summary
In the current construction process of water conservancy projects, the manual supervision model suffers from problems such as low efficiency, blind spots in supervision, and high human resource costs, making it difficult to achieve efficient and comprehensive safety monitoring.
A target water conservancy project safety monitoring model is carried out using a drone. By acquiring sample weather parameters and shooting distance, a deep learning model is trained to determine shooting locations. Construction images are captured in real time and input into the model for safety monitoring. Convolutional attention layers and spatial pooling layers are combined for feature extraction and prediction to achieve efficient and accurate safety monitoring.
It has enabled efficient and accurate safety monitoring of water conservancy projects, reduced human resource costs, improved monitoring coverage and regulatory efficiency, and reduced the risk of regulatory blind spots.
Smart Images
Figure CN121545114A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of safety monitoring of water conservancy projects, and in particular to a method, device and storage medium for safety monitoring of water conservancy projects. Background Technology
[0002] Water conservancy engineering, as a systematic engineering technology field involving the comprehensive development, utilization, protection, and management of water resources, encompasses multiple key aspects, from preliminary surveying, scientific planning, and engineering design to specific construction, post-construction maintenance, and scientific research and innovation. The main goal of this engineering technology system is to effectively address a series of major water-related issues concerning people's livelihoods, such as water resource development and utilization, flood control and disaster reduction, and water ecological protection. Due to the crucial role of water conservancy projects in national infrastructure construction and people's livelihood security, a major safety accident during construction can have extremely serious chain reactions. Such hazards not only directly threaten the lives and property of surrounding residents but may also have long-term negative impacts on regional economic development and ecological environment stability. Under current technological conditions, safety supervision during the construction process of water conservancy projects mainly relies on manual inspections and on-site monitoring. However, because water conservancy projects are typically large-scale, widely distributed, and have long construction periods, this traditional manual supervision model faces many challenges. On the one hand, ensuring comprehensive supervision requires a large number of professional and technical personnel, resulting in high human resource costs; on the other hand, manual supervision inevitably suffers from blind spots and negligence due to fatigue. These factors together have led to significant technical shortcomings in the current safety supervision of water conservancy projects, such as low efficiency and delayed response, which urgently need to be improved and perfected through technological innovation. Summary of the Invention
[0003] Based on this, the purpose of this application is to provide a method, device and storage medium for safety monitoring of water conservancy projects, which can overcome the shortcomings of the prior art.
[0004] To achieve the above objectives, the technical solution adopted in this application is as follows:
[0005] The first aspect of this application provides a method for safety monitoring of water conservancy projects, including:
[0006] The system acquires several target water conservancy project safety monitoring models that have a monitoring accuracy higher than a preset accuracy threshold for several types of sample weather parameters and have the largest shooting distance; the target water conservancy project safety monitoring models are used to identify the water conservancy project safety monitoring results based on construction images;
[0007] Based on the weather parameters of the water conservancy project to be monitored, obtain the corresponding safety monitoring model of the target water conservancy project, as well as the maximum shooting distance corresponding to the safety monitoring model of the target water conservancy project;
[0008] The shooting locations are obtained based on the maximum shooting distance and the location information of the water conservancy project to be monitored;
[0009] Drive the drone to the shooting location to capture real-time construction images of the water conservancy project to be monitored;
[0010] The real-time construction images are input into the target water conservancy project safety monitoring model to obtain the water conservancy project safety monitoring results.
[0011] Compared with traditional technologies, the beneficial effects of this application are:
[0012] The water conservancy project safety monitoring method of this application obtains several target water conservancy project safety monitoring models with monitoring accuracy higher than a preset accuracy threshold for several types of sample weather parameters and the largest shooting distance. Based on the weather parameters of the water conservancy project to be monitored, the method obtains the corresponding target water conservancy project safety monitoring model and the maximum shooting distance corresponding to the target water conservancy project safety monitoring model. Then, based on the maximum shooting distance and the location information of the water conservancy project to be monitored, the method obtains the shooting point. The method then drives a drone to fly to the shooting point to capture real-time construction images of the water conservancy project to be monitored. Finally, the method inputs the real-time construction images into the target water conservancy project safety monitoring model to obtain the water conservancy project safety monitoring results. This method can efficiently and accurately achieve water conservancy project safety monitoring.
[0013] As one implementation method, the step of acquiring several target water conservancy project safety monitoring models whose monitoring accuracy for corresponding weather parameters is higher than a preset accuracy threshold and whose shooting distance is the largest includes:
[0014] Obtain several sets of water conservancy project image samples corresponding to several types of sample weather parameters; each set of water conservancy project image samples includes several water conservancy project safety monitoring image samples; the water conservancy project safety monitoring image samples include construction images labeled with safety monitoring data;
[0015] Based on the shooting distance, the several water conservancy project safety monitoring image samples in each of the water conservancy project image sample sets are grouped to obtain several shooting distance image sample groups corresponding to various sample weather parameters.
[0016] A deep learning model is trained based on several sets of image samples taken at various shooting distances for various sample weather parameters, resulting in several initial water conservancy project safety monitoring models corresponding to several sets of image samples taken at various shooting distances for various sample weather parameters.
[0017] Based on the monitoring accuracy of the aforementioned initial water conservancy project safety monitoring models and the corresponding shooting distance of the image sample groups, a target water conservancy project safety monitoring model corresponding to various sample weather parameters is obtained.
[0018] In this embodiment, after grouping the several water conservancy project safety monitoring image samples of each water conservancy project image sample set according to the shooting distance, a deep learning model is trained according to several shooting distance image sample groups corresponding to various sample weather parameters to obtain several shooting distance image sample groups. Combining the monitoring accuracy and the shooting distance of the corresponding required image, the target water conservancy project safety monitoring model corresponding to various sample weather parameters can be accurately obtained.
[0019] As one implementation method, the step of obtaining the target water conservancy project safety monitoring model corresponding to various sample weather parameters based on the monitoring accuracy of the plurality of initial water conservancy project safety monitoring models and the shooting distance of the corresponding shooting distance image sample group includes:
[0020] The monitoring accuracy of the initial water conservancy project safety monitoring models corresponding to various sample weather parameters is compared with the accuracy threshold to obtain several candidate water conservancy project safety monitoring models that are greater than the accuracy threshold for various sample weather parameters.
[0021] Based on the shooting distance of the shooting distance image sample group corresponding to the candidate water conservancy project safety monitoring models with the same sample weather parameters, the target water conservancy project safety monitoring model with the largest shooting distance corresponding to various sample weather parameters is obtained from the candidate water conservancy project safety monitoring models.
[0022] In this embodiment, after obtaining several candidate water conservancy project safety monitoring models that are greater than the accuracy threshold for various sample weather parameters, the target water conservancy project safety monitoring model with the largest shooting distance for various sample weather parameters can be effectively obtained based on the shooting distance.
[0023] In one implementation method, the location information of the water conservancy project to be monitored includes the locations of several construction sites; the shooting points include a first shooting point and a second shooting point.
[0024] The step of obtaining the shooting location based on the maximum shooting distance and the location information of the water conservancy project to be monitored includes:
[0025] Using the locations of the aforementioned construction sites as the center and the maximum shooting distance as the radius, several shooting ranges corresponding to the aforementioned construction sites are obtained;
[0026] Based on the plurality of shooting ranges that have overlapping ranges, a first shooting point is obtained; the first shooting point is a common shooting point of the plurality of construction sites of the plurality of shooting ranges that have overlapping ranges.
[0027] The step of driving the drone to fly to the shooting location to capture real-time construction images of the water conservancy project to be monitored includes:
[0028] The drone is driven to fly to the first shooting point to capture the first real-time construction image; the first real-time construction image is a real-time construction image of several construction sites corresponding to the intersection range covered by the shooting range.
[0029] In this embodiment, the drone is driven to the first shooting point so that the drone can simultaneously capture a first real-time construction image including real-time construction images of several construction sites, thereby improving the efficiency of acquiring real-time construction images.
[0030] In one implementation, the shooting location includes a second shooting location;
[0031] The step of obtaining the shooting location based on the maximum shooting distance and the location information of the water conservancy project to be monitored includes:
[0032] Based on the shooting ranges where there is no overlap and the preset shooting distance, obtain the corresponding second shooting point;
[0033] The step of driving the drone to fly to the shooting location to capture real-time construction images of the water conservancy project to be monitored includes:
[0034] The drone is driven to fly to the second shooting point to capture second real-time construction images of several construction sites that do not overlap.
[0035] In this embodiment, driving the drone to the second shooting point to capture the second real-time construction image can obtain real-time construction images of several construction sites that do not overlap with each other on a one-to-one basis.
[0036] As one implementation method, the target water conservancy project safety monitoring model includes a backbone network, a neck network, and a head network connected in sequence; the backbone network includes convolutional layers, multiple convolutional attention layers, and spatial pooling layers;
[0037] The step of inputting the real-time construction image into the target water conservancy project safety monitoring model to obtain the water conservancy project safety monitoring results includes:
[0038] The real-time construction image is input into the convolutional layer for feature extraction to obtain the first sample features;
[0039] The first sample features are input into the cascaded convolutional attention layers for feature extraction and feature attention processing to obtain the second sample features output by each convolutional attention layer;
[0040] The second sample feature output from the last convolutional attention layer is input into the spatial pooling layer for dimensionality reduction to obtain the pooled feature.
[0041] The pooling features and several preset second sample features are input into the neck network for fusion processing and spatial transformation attention processing to obtain multiple fused attention features;
[0042] The multiple fused attention features are input into the head network for prediction processing to obtain the safety monitoring results of the water conservancy project.
[0043] In this embodiment, the real-time construction image is input into the convolutional layer for feature extraction to obtain first sample features; the first sample features are input into the cascaded convolutional attention layers for feature extraction and feature attention processing to obtain second sample features output by each convolutional attention layer; the second sample features output by the last convolutional attention layer are input into the spatial pooling layer for dimensionality reduction to obtain pooling features; the pooling features and several preset second sample features are input into the neck network for fusion processing and spatial transformation attention processing to obtain multiple fused attention features; the multiple fused attention features are input into the head network for prediction processing, which can accurately obtain the safety monitoring results of water conservancy projects.
[0044] In one implementation, the neck network includes a first upsampling fusion module, a first transform attention module, a second upsampling fusion module, a second transform attention module, a first convolutional fusion module, a third transform attention module, a second convolutional fusion module, and a fourth transform attention module;
[0045] After the pooling feature is upsampled by the first upsampling fusion module, it is fused with the second sample feature output by the penultimate convolutional attention layer to obtain the first upsampling fusion feature.
[0046] The first transform attention module performs transform attention processing on the first upsampled fused features to obtain the first transform attention features;
[0047] The second upsampling fusion module upsamples the first transform attention feature and then fuses it with the second sample feature output by the third-to-last convolutional attention layer to obtain the second upsampling fusion feature.
[0048] The second transform attention module performs transform attention processing on the first transform attention features to obtain the second transform attention features;
[0049] After performing convolution processing on the second transform attention feature, the first convolution fusion module fuses it with the first transform attention feature to obtain the first convolution fusion feature;
[0050] The third transform attention module performs transform attention processing on the first convolutional fusion feature to obtain the third transform attention feature;
[0051] The second convolutional fusion module performs convolution processing on the three-transformation attention features and then fuses them with the pooling features to obtain the second convolutional fusion features;
[0052] The fourth transform attention module performs transform attention processing on the second convolutional fusion feature to obtain the fourth transform attention feature;
[0053] The second transform attention feature, the third transform attention feature, and the fourth transform attention feature are determined as the fused attention feature.
[0054] In this embodiment, by performing fusion and transformation attention processing on features of different spatial dimensions through a first upsampling fusion module, a first transform attention module, a second upsampling fusion module, a second transform attention module, a first convolutional fusion module, a third transform attention module, a second convolutional fusion module, and a fourth transform attention module, the neck network's ability to capture features of real-time construction images of different spatial dimensions can be enhanced.
[0055] In one implementation, the head network includes multiple decoupling heads, each of which corresponds one-to-one with a fusion attention feature, and is used to predict the safety monitoring results of the water conservancy project based on the corresponding fusion attention feature.
[0056] In this embodiment, the safety monitoring results of the water conservancy project can be accurately obtained through multiple decoupling heads.
[0057] A second aspect of this application provides a water conservancy project safety monitoring device, comprising:
[0058] The target monitoring model acquisition module acquires several target water conservancy project safety monitoring models whose monitoring accuracy for several types of sample weather parameters is higher than a preset accuracy threshold and whose shooting distance is the largest; the target water conservancy project safety monitoring models are used to identify the water conservancy project safety monitoring results based on construction images;
[0059] The maximum shooting distance acquisition module is used to acquire the corresponding target water conservancy project safety monitoring model and the maximum shooting distance corresponding to the target water conservancy project safety monitoring model based on the weather parameters of the water conservancy project to be monitored.
[0060] The shooting location acquisition module is used to obtain the shooting location based on the maximum shooting distance and the location information of the water conservancy project to be monitored;
[0061] A real-time construction image capturing module is used to drive a drone to fly to the capturing point to capture real-time construction images of the water conservancy project to be monitored;
[0062] The water conservancy project safety monitoring module is used to input the real-time construction images into the target water conservancy project safety monitoring model to obtain the water conservancy project safety monitoring results of the water conservancy project to be monitored.
[0063] A third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the water conservancy project safety monitoring method described above.
[0064] To better understand and implement this application, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description
[0065] Figure 1 This is a flowchart of a water conservancy project safety monitoring method according to an embodiment of this application;
[0066] Figure 2 This is a schematic diagram of the structure of a target water conservancy project safety monitoring model according to an embodiment of this application;
[0067] Figure 3 This is a schematic diagram of the module connection of a water conservancy project safety monitoring device according to an embodiment of this application.
[0068] 100. Water conservancy project safety monitoring device; 101. Target monitoring model acquisition module; 102. Maximum shooting distance acquisition module; 103. Shooting point acquisition module; 104. Real-time construction image shooting module; 105. Water conservancy project safety monitoring module. Detailed Implementation
[0069] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.
[0070] It should be understood that the described embodiments are merely some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of the embodiments of this application.
[0071] In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. In the description of this application, it should be understood that the terms "first," "second," "third," etc., are used only to distinguish similar objects and are not necessarily used to describe a specific order or sequence, nor should they be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances. The singular forms "a," "the," and "the" used in this application and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. The word "if" as used herein can be interpreted as "when," "when," or "in response to determination."
[0072] Furthermore, in the description of this application, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.
[0073] Please see Figure 1 The flowchart of the water conservancy project safety monitoring method according to the first embodiment of this application includes:
[0074] S1: Obtain several target water conservancy project safety monitoring models with monitoring accuracy higher than a preset accuracy threshold and the largest shooting distance for several types of sample weather parameters; the target water conservancy project safety monitoring models are used to identify the water conservancy project safety monitoring results based on construction images;
[0075] S2: Based on the weather parameters of the water conservancy project to be monitored, obtain the corresponding safety monitoring model of the target water conservancy project and the maximum shooting distance corresponding to the safety monitoring model of the target water conservancy project;
[0076] S3: Based on the maximum shooting distance and the location information of the water conservancy project to be monitored, the shooting points are obtained;
[0077] S4: Drive the drone to the shooting location to capture real-time construction images of the water conservancy project to be monitored;
[0078] S5: Input the real-time construction image into the target water conservancy project safety monitoring model to obtain the water conservancy project safety monitoring results of the water conservancy project to be monitored.
[0079] Compared with traditional technologies, the beneficial effects of this application are:
[0080] The water conservancy project safety monitoring method of this application obtains several target water conservancy project safety monitoring models with monitoring accuracy higher than a preset accuracy threshold for several types of sample weather parameters and the largest shooting distance. Based on the weather parameters of the water conservancy project to be monitored, the method obtains the corresponding target water conservancy project safety monitoring model and the maximum shooting distance corresponding to the target water conservancy project safety monitoring model. Then, based on the maximum shooting distance and the location information of the water conservancy project to be monitored, the method obtains the shooting point. The method then drives a drone to fly to the shooting point to capture real-time construction images of the water conservancy project to be monitored. Finally, the method inputs the real-time construction images into the target water conservancy project safety monitoring model to obtain the water conservancy project safety monitoring results. This method can efficiently and accurately achieve water conservancy project safety monitoring.
[0081] In a feasible embodiment, step S1: obtaining several target water conservancy project safety monitoring models whose monitoring accuracy for several types of sample weather parameters is higher than a preset accuracy threshold and whose shooting distance is the largest, includes:
[0082] S11: Obtain several sets of water conservancy project image samples corresponding to several types of sample weather parameters; each set of water conservancy project image samples includes several water conservancy project safety monitoring image samples; the water conservancy project safety monitoring image samples include construction images labeled with safety monitoring data;
[0083] S12: Based on the shooting distance, group the several water conservancy project safety monitoring image samples in each of the water conservancy project image sample sets to obtain several shooting distance image sample groups corresponding to various sample weather parameters;
[0084] S13: Train a deep learning model based on the several shooting distance image sample groups of various sample weather parameters to obtain several initial water conservancy project safety monitoring models corresponding to the several shooting distance image sample groups of various sample weather parameters.
[0085] S14: Based on the monitoring accuracy of the several initial water conservancy project safety monitoring models and the shooting distance of the corresponding shooting distance image sample group, obtain the target water conservancy project safety monitoring model corresponding to various sample weather parameters.
[0086] In this embodiment, after grouping the several water conservancy project safety monitoring image samples of each water conservancy project image sample set according to the shooting distance, a deep learning model is trained according to several shooting distance image sample groups corresponding to various sample weather parameters to obtain several shooting distance image sample groups. Combining the monitoring accuracy and the shooting distance of the corresponding required image, the target water conservancy project safety monitoring model corresponding to various sample weather parameters can be accurately obtained.
[0087] In a feasible embodiment, step S14, which involves obtaining the target water conservancy project safety monitoring model corresponding to various sample weather parameters based on the monitoring accuracy of the plurality of initial water conservancy project safety monitoring models and the shooting distance of the corresponding shooting distance image sample group, includes:
[0088] S141: Compare the monitoring accuracy of the several initial water conservancy project safety monitoring models corresponding to various sample weather parameters with the accuracy threshold to obtain several candidate water conservancy project safety monitoring models that are greater than the accuracy threshold corresponding to various sample weather parameters.
[0089] S142: Based on the shooting distance of the shooting distance image sample group corresponding to the candidate water conservancy project safety monitoring models with the same sample weather parameters, obtain the target water conservancy project safety monitoring model with the largest shooting distance corresponding to various sample weather parameters from the candidate water conservancy project safety monitoring models.
[0090] In this embodiment, after obtaining several candidate water conservancy project safety monitoring models that are greater than the accuracy threshold for various sample weather parameters, the target water conservancy project safety monitoring model with the largest shooting distance for various sample weather parameters can be effectively obtained based on the shooting distance.
[0091] In one feasible embodiment, the location information of the water conservancy project to be monitored includes the locations of several construction sites; the shooting points include a first shooting point and a second shooting point;
[0092] Step S3: The step of obtaining the shooting location based on the maximum shooting distance and the location information of the water conservancy project to be monitored includes:
[0093] S31: Using the locations of the aforementioned construction sites as the center and the maximum shooting distance as the radius, obtain several shooting ranges corresponding to the aforementioned construction sites;
[0094] S32: Based on the plurality of shooting ranges that have overlapping ranges, obtain a first shooting point; the first shooting point is a common shooting point of the plurality of construction sites of the plurality of shooting ranges that have overlapping ranges;
[0095] S4: The step of driving the drone to fly to the shooting location to capture real-time construction images of the water conservancy project to be monitored includes:
[0096] S41: Drive the drone to fly to the first shooting point and shoot the first real-time construction image; the first real-time construction image is a real-time construction image of several construction sites corresponding to the intersection range covered by the shooting range.
[0097] If the required shooting wide-angle of several construction sites corresponding to the intersection range of the first shooting point is greater than the shooting wide-angle parameter of the camera installed on the drone, the second shooting point is obtained again based on the shooting range corresponding to the several construction sites and the preset shooting distance.
[0098] In this embodiment, the drone is driven to the first shooting point so that it can simultaneously capture a first real-time construction image including real-time construction images of several construction sites, thereby improving the efficiency of acquiring real-time construction images.
[0099] In one feasible embodiment, the shooting location includes a second shooting location;
[0100] The step of obtaining the shooting location based on the maximum shooting distance and the location information of the water conservancy project to be monitored includes:
[0101] Based on the shooting ranges where there is no overlap and the preset shooting distance, obtain the corresponding second shooting point;
[0102] S4: The step of driving the drone to fly to the shooting location to capture real-time construction images of the water conservancy project to be monitored includes:
[0103] S42: Drive the drone to the second shooting point to capture a second real-time construction image of several construction sites that do not overlap.
[0104] In this embodiment, driving the drone to the second shooting point to capture the second real-time construction image can obtain real-time construction images of several construction sites that do not overlap with each other on a one-to-one basis.
[0105] Please see Figure 2 In one feasible embodiment, the target water conservancy project safety monitoring model includes a backbone network, a neck network, and a head network connected in sequence; the backbone network includes convolutional layers (Conv), multiple convolutional attention layers (Conv and DAAM), and spatial pooling layers (BSPPF);
[0106] Step S5: The step of inputting the real-time construction image into the target water conservancy project safety monitoring model to obtain the water conservancy project safety monitoring results includes:
[0107] S51: Input the real-time construction image into the convolutional layer for feature extraction to obtain the first sample features;
[0108] S52: Input the first sample features into the cascaded multiple convolutional attention layers for feature extraction and feature attention processing to obtain the second sample features output by each convolutional attention layer;
[0109] S53: Input the second sample feature output from the last convolutional attention layer into the spatial pooling layer for dimensionality reduction to obtain pooled features;
[0110] S54: Input the pooling features and several preset second sample features into the neck network for fusion processing and spatial transformation attention processing to obtain multiple fused attention features;
[0111] S55: Input the multiple fused attention features into the head network for prediction processing to obtain the safety monitoring results of the water conservancy project.
[0112] In this embodiment, the real-time construction image is input into the convolutional layer for feature extraction to obtain the first sample feature; the first sample feature is input into the cascaded multiple convolutional attention layers for feature extraction and feature attention processing to obtain the second sample feature output by each convolutional attention layer; the second sample feature output by the last convolutional attention layer is input into the spatial pooling layer for dimensionality reduction processing to obtain the pooling feature; the pooling feature and several preset second sample features are input into the neck network for fusion processing and spatial transformation attention processing to obtain multiple fused attention features; the multiple fused attention features are input into the head network for prediction processing, which can accurately obtain the safety monitoring results of the water conservancy project.
[0113] In one feasible embodiment, the neck network includes a first upsample fusion module (Upsample and Concat), a first transform attention module (VSSA), a second upsample fusion module (Upsample and Concat), a second transform attention module (VSSA), a first convolutional fusion module (Conv and Concat), a third transform attention module (VSSA), a second convolutional fusion module (Conv and Concat), and a fourth transform attention module (VSSA);
[0114] After the pooling feature is upsampled by the first upsampling fusion module, it is fused with the second sample feature output by the penultimate convolutional attention layer to obtain the first upsampling fusion feature.
[0115] The first transform attention module performs transform attention processing on the first upsampled fused features to obtain the first transform attention features;
[0116] The second upsampling fusion module upsamples the first transform attention feature and then fuses it with the second sample feature output by the third-to-last convolutional attention layer to obtain the second upsampling fusion feature.
[0117] The second transform attention module performs transform attention processing on the first transform attention features to obtain the second transform attention features;
[0118] After performing convolution processing on the second transform attention feature, the first convolution fusion module fuses it with the first transform attention feature to obtain the first convolution fusion feature;
[0119] The third transform attention module performs transform attention processing on the first convolutional fusion feature to obtain the third transform attention feature;
[0120] The second convolutional fusion module performs convolution processing on the three-transformation attention features and then fuses them with the pooling features to obtain the second convolutional fusion features;
[0121] The fourth transform attention module performs transform attention processing on the second convolutional fusion feature to obtain the fourth transform attention feature;
[0122] The second transform attention feature, the third transform attention feature, and the fourth transform attention feature are determined as the fused attention feature.
[0123] In this embodiment, by performing fusion and transformation attention processing on features of different spatial dimensions through a first upsampling fusion module, a first transform attention module, a second upsampling fusion module, a second transform attention module, a first convolutional fusion module, a third transform attention module, a second convolutional fusion module, and a fourth transform attention module, the neck network's ability to capture features of real-time construction images of different spatial dimensions can be enhanced.
[0124] In one feasible embodiment, the head network includes multiple decoupling heads (Detect), each decoupling head corresponding to a fusion attention feature, and is used to predict the safety monitoring results of the water conservancy project based on the corresponding fusion attention feature.
[0125] Specifically, the decoupling head includes a regression branch module and a classification branch module. The regression branch module has a loss function to calculate the positional offset between the predicted bounding box of the regression branch and the detection label of the marine organism training sample. The classification branch model obtains the value of each position on the classifier output tensor through pooling and convolution operations for the predicted bounding box, which represents the probability that the predicted bounding box belongs to each category. Finally, the final detection result is selected through maximum suppression.
[0126] In this embodiment, the safety monitoring results of the water conservancy project can be accurately obtained through multiple decoupling heads.
[0127] Please see Figure 3 The second embodiment of this application provides a water conservancy project safety monitoring device 100, including:
[0128] The target water conservancy project safety monitoring model acquisition module acquires several target water conservancy project safety monitoring models whose monitoring accuracy for corresponding weather parameters is higher than a preset accuracy threshold and whose shooting distance is the largest; the target water conservancy project safety monitoring model is used to identify the water conservancy project safety monitoring results based on construction images;
[0129] The maximum shooting distance acquisition module 102 is used to acquire the corresponding target water conservancy project safety monitoring model and the maximum shooting distance corresponding to the target water conservancy project safety monitoring model based on the weather parameters of the water conservancy project to be monitored.
[0130] The shooting point acquisition module 103 is used to obtain the shooting point based on the maximum shooting distance and the location information of the water conservancy project to be monitored;
[0131] The real-time construction image capturing module 104 is used to drive the drone to fly to the capturing point to capture real-time construction images of the water conservancy project to be monitored.
[0132] The water conservancy project safety monitoring module 105 is used to input the real-time construction images into the target water conservancy project safety monitoring model to obtain the water conservancy project safety monitoring results of the water conservancy project to be monitored.
[0133] It should be noted that the water conservancy project safety monitoring device 100 provided in the second embodiment of this application is only illustrated by the above-described division of functional modules when performing the water conservancy project safety monitoring method. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the water conservancy project safety monitoring device 100 provided in the second embodiment of this application and the water conservancy project safety monitoring method of the first embodiment of this application belong to the same concept, and its implementation process is detailed in the method embodiment, which will not be repeated here.
[0134] A third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the water conservancy project safety monitoring method described above.
[0135] The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without any inventive effort.
[0136] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0137] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function selected in one or more boxes.
[0138] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function selected in one or more boxes.
[0139] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0140] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0141] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0142] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0143] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.
Claims
1. A method of monitoring the safety of a hydraulic structure, characterized by, The method comprises the following steps: obtaining a plurality of target water conservancy project safety monitoring models corresponding to a plurality of sample weather parameters and having a monitoring accuracy higher than a preset accuracy threshold and a maximum shooting distance; the target water conservancy project safety monitoring model is used to identify a water conservancy project safety monitoring result according to a construction image; obtaining a target water conservancy project safety monitoring model corresponding to a weather parameter of a water conservancy project to be monitored and a maximum shooting distance corresponding to the target water conservancy project safety monitoring model; obtaining a shooting point according to the maximum shooting distance and position information of the water conservancy project to be monitored; driving a drone to fly to the shooting point to shoot a real-time construction image of the water conservancy project to be monitored; inputting the real-time construction image into the target water conservancy project safety monitoring model to obtain a water conservancy project safety monitoring result of the water conservancy project to be monitored.
2. The hydraulic engineering safety monitoring method according to claim 1, characterized in that, The step of obtaining a plurality of target water conservancy project safety monitoring models corresponding to a plurality of sample weather parameters and having a monitoring accuracy higher than a preset accuracy threshold and a maximum shooting distance comprises the following steps: obtaining a plurality of water conservancy project image sample sets corresponding to a plurality of sample weather parameters; each water conservancy project image sample set comprises a plurality of water conservancy project safety monitoring image samples; the water conservancy project safety monitoring image sample comprises a construction image labeled with safety monitoring data; grouping the plurality of water conservancy project safety monitoring image samples of each water conservancy project image sample set according to a shooting distance to obtain a plurality of shooting distance image sample groups corresponding to various sample weather parameters; training a deep learning model according to the plurality of shooting distance image sample groups corresponding to various sample weather parameters to obtain a plurality of initial water conservancy project safety monitoring models of the plurality of shooting distance image sample groups corresponding to various sample weather parameters; obtaining target water conservancy project safety monitoring models corresponding to various sample weather parameters according to the monitoring accuracy of the plurality of initial water conservancy project safety monitoring models and the shooting distance of the corresponding shooting distance image sample groups.
3. The hydraulic-engineering safety monitoring method according to claim 2, characterized in that, The step of obtaining target water conservancy project safety monitoring models corresponding to various sample weather parameters according to the monitoring accuracy of the plurality of initial water conservancy project safety monitoring models and the shooting distance of the corresponding shooting distance image sample groups comprises the following steps: comparing the monitoring accuracy of the plurality of initial water conservancy project safety monitoring models corresponding to various sample weather parameters with an accuracy threshold to obtain a plurality of candidate water conservancy project safety monitoring models corresponding to various sample weather parameters and greater than the accuracy threshold; obtaining the target water conservancy project safety monitoring model corresponding to various sample weather parameters and having a maximum shooting distance from the candidate water conservancy project safety monitoring models according to the shooting distance of the shooting distance image sample group corresponding to the candidate water conservancy project safety monitoring model of the same sample weather parameter.
4. The hydraulic engineering safety monitoring method according to claim 1, characterized in that, The position information of the water conservancy project to be monitored comprises the location of a plurality of construction sites; the shooting point comprises a first shooting point and a second shooting point. The step of obtaining a shooting point according to the maximum shooting distance and the position information of the water conservancy project to be monitored comprises the following steps: The maximum shooting distance is a radius, and a plurality of shooting ranges corresponding to the plurality of construction sites are obtained; A first shooting point is obtained according to the plurality of shooting ranges with the intersection range, and the first shooting point is a common shooting point of the plurality of construction sites with the intersection range; The step of driving the unmanned aerial vehicle to fly to the shooting point to shoot the real-time construction image of the water conservancy project to be monitored comprises: The unmanned aerial vehicle is driven to fly to the first shooting point to shoot a first real-time construction image, and the first real-time construction image is a real-time construction image of a plurality of construction sites corresponding to the intersection range covered by the shooting range.
5. The hydraulic engineering safety monitoring method according to claim 1, wherein The shooting point comprises a second shooting point; The step of obtaining the shooting point according to the maximum shooting distance and the position information of the water conservancy project to be monitored comprises: A second shooting point corresponding to each shooting range without the intersection range and a preset shooting distance is obtained; The step of driving the unmanned aerial vehicle to fly to the shooting point to shoot the real-time construction image of the water conservancy project to be monitored comprises: The unmanned aerial vehicle is driven to fly to the second shooting point to shoot a second real-time construction image of a plurality of construction sites without the intersection range.
6. The hydraulic engineering safety monitoring method according to claim 1, wherein The target water conservancy project safety monitoring model comprises a main network, a neck network and a head network connected in sequence; the main network comprises a convolution layer, a plurality of convolution attention layers and a spatial pooling layer; The step of inputting the real-time construction image into the target water conservancy project safety monitoring model to obtain the water conservancy project safety monitoring result of the water conservancy project to be monitored comprises: The real-time construction image is input into the convolution layer for feature extraction to obtain a first sample feature; The first sample feature is input into the plurality of convolution attention layers connected in cascade for feature extraction and feature attention processing to obtain a second sample feature output by each convolution attention layer; The second sample feature output by the last convolution attention layer is input into the spatial pooling layer for dimension reduction processing to obtain a pooled feature; The pooled feature and a plurality of preset second sample features are input into the neck network for fusion processing and spatial transformation attention processing to obtain a plurality of fusion attention features; The plurality of fusion attention features are input into the head network for prediction processing to obtain the water conservancy project safety monitoring result.
7. The hydraulic-engineering safety monitoring method according to claim 6, characterized in that, The neck network comprises a first upsampling fusion module, a first transformation attention module, a second upsampling fusion module, a second transformation attention module, a first convolution fusion module, a third transformation attention module, a second convolution fusion module and a fourth transformation attention module; The first upsampling fusion module performs upsampling processing on the pooled feature, and then performs fusion processing on the second sample feature output by the second-to-last convolution attention layer to obtain a first upsampling fusion feature; The first transformation attention module performs transformation attention processing on the first upsampling fusion feature to obtain a first transformation attention feature; The first transformation attention module performs transformation attention processing on the first upsampling fusion feature to obtain a first transformation attention feature; The second upsampling fusion module fuses the first transformed attention feature with a second sample feature output by a third-to-last convolutional attention layer after performing upsampling processing on the first transformed attention feature, to obtain a second upsampling fusion feature; The second transformed attention module performs transformed attention processing on the first transformed attention feature to obtain a second transformed attention feature; The first convolutional fusion module fuses the second transformed attention feature with the first transformed attention feature after performing convolutional processing on the second transformed attention feature, to obtain a first convolutional fusion feature; The third transformed attention module performs transformed attention processing on the first convolutional fusion feature to obtain a third transformed attention feature; The second convolutional fusion module fuses the third transformed attention feature with the pooled feature after performing convolutional processing on the third transformed attention feature, to obtain a second convolutional fusion feature; The fourth transformed attention module performs transformed attention processing on the second convolutional fusion feature to obtain a fourth transformed attention feature; The second transformed attention feature, the third transformed attention feature, and the fourth transformed attention feature are determined as the fusion attention features.
8. The hydraulic engineering safety monitoring method according to claim 6, wherein, The head network includes a plurality of decoupled heads, each of which corresponds to each fusion attention feature one-to-one, and is configured to perform prediction learning processing according to the corresponding fusion attention feature.
9. A hydraulic engineering safety monitoring device, characterized by, The method comprises: The target water conservancy project safety monitoring model acquisition module acquires a plurality of target water conservancy project safety monitoring models corresponding to a plurality of sample weather parameters, the monitoring accuracy of which is higher than a preset accuracy threshold, and the shooting distance of which is the largest; the target water conservancy project safety monitoring model is used to identify a water conservancy project safety monitoring result according to a construction image; The maximum shooting distance acquisition module is configured to acquire a target water conservancy project safety monitoring model corresponding to a weather parameter of a water conservancy project to be monitored, and a maximum shooting distance corresponding to the target water conservancy project safety monitoring model; The shooting point acquisition module is configured to acquire a shooting point according to the maximum shooting distance and position information of the water conservancy project to be monitored; The real-time construction image shooting module is configured to drive a drone to fly to the shooting point to shoot a real-time construction image of the water conservancy project to be monitored; The water conservancy project safety monitoring module is configured to input the real-time construction image into the target water conservancy project safety monitoring model to obtain a water conservancy project safety monitoring result of the water conservancy project to be monitored.
10. A computer-readable storage medium, the computer-readable storage medium storing a computer program, characterized in that: The computer program, when executed by a processor, implements the steps of the water conservancy project safety monitoring method of any one of claims 1 to 8.